UWB Radar Signal Kick Detection for Tailgate Unlocking Based on Spatio-Temporal Network

Yuhan Wang, Shuang Yan, Yiling Fu, Tianyun Shangguan, Yuxuan Sun, Lin Sun

Abstract: Inductive tailgate systems have significantly enhanced user convenience in vehicle operations. However, the performance of current systems can be hindered by environmental factors, leading to reduced recognition accuracy and durability over time. To address these issues, we propose a kick gesture recognition method based on ultra-wideband (UWB) signals, implemented through a spatial-temporal neural network. Our method starts with employing a Time-Range-Doppler method to extract kick gesture sequences that contain temporal and spatial features from preprocessed data. These sequences are then fed into a network containing both Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) cells to predict the final results, where the CNN section is used to learn spatial features and the RNN section is used to learn temporal features. To verify the effectiveness of our approach, we conducted experiments in three typical parking environments: underground garages, overgrown lawns, and sloped terrains, collecting data from 12 participants for model training and testing. Experimental results demonstrated the feasibility of using UWB signals for wireless kick gesture recognition and confirmed the high accuracy of our proposed method.